Appeal courts: gender, convictions and acquittals

Research checked on 1 November 2026. Official sanction totals are available. A table crossing appeal-court guilt outcomes with defendants’ sex has not been found in the sources examined. Nor do those sources provide outcomes by the sex of the judicial panel. The requested rate cannot responsibly be calculated from those sources. This is a data gap, not evidence that gender has no effect.

What the existing official figures measure

Statistics Norway table 13725 counts registered criminal sanctions with a court of appeal as the court instance. These are not unique defendants, all criminal appeals or all guilt determinations. A person may have multiple sanctions. Registration year is not necessarily the year judgment was delivered.

The API extract was retrieved on 1 November 2026. Its latest complete year is 2025, updated 2 July 2026. Table 13725 has no sex dimension or acquittal category. Table 13724 has sex but no court-instance dimension. Its 2025 totals of 193,994 sanctions for men and 62,273 for women include other instances and types of sanction, including fines. They cannot be used as counts of people convicted in appeal courts. The national sex distribution cannot be applied to a separate appeal-court total to manufacture a cross-table.

The Norwegian Courts Administration's 2025 annual report, page 18, gives 2,509 decided appeal-screening cases, 427 decided evidentiary appeals with a maximum statutory sentence of six years or less, and 253 above six years. These are national case-management counts, not defendant-case outcomes. Screening and a later full hearing are different stages; adding them can count the same matter twice. Even the 680 evidentiary-appeal cases do not supply a ready denominator by defendants’ sex.

New source: Gulating’s own 2025 annual report

The actual report has now been retrieved from Gulating’s official publication index, rather than using the national report as a substitute. Physical PDF page 5 (printed page 4) lists 32 permanent judicial posts and states that approximately 40% of the judges were women. This is staffing information, not a comparison of judicial decisions by gender.

Physical PDF page 13 (printed page 12) supplies these case counts. The mixed-bench category includes both evidentiary appeals below six years and limited mixed-bench appeals. Therefore 47 + 76 is not automatically a count of all guilt appeals. The report does not provide the requested guilt-outcome-by-sex cross-table.

Gulating annual report 2025 ↗ · Download the bounded context figures (CSV)

Three different research questions

1. Defendants: do men and women have different acquittal rates in comparable guilt appeals? This requires counts of acquittals and convictions within the same defined population.
2. Judicial composition: are outcomes different across different professional/lay-judge panels? This requires the whole panel, eligible case selection and case characteristics, not an assumed gender inferred from names or photographs.
3. Sentencing: do already-convicted defendants receive different sanctions? This is a different question and cannot answer an acquittal-rate question.

Statistics Norway describes sex as a register variable, mainly from the Population Register. It should not automatically be described as self-reported gender identity. No person’s sex or gender is inferred by this site.

A collective decision cannot be assigned to one judge

Section 35 of the Criminal Procedure Act requires five votes for an appeal-court guilt decision against the accused, at least one from a professional judge. Participation in a panel does not establish a particular judge’s individual position. A raw rate also does not establish bias, unlawful conduct or financial motivation: differences in case allocation, offences, evidence and appeal selection may matter.

The previous legal-source retrieval obtained the provision’s text through Legal Data Hunter/Lovdata, but citation resolution failed and the provider mislabelled some Norwegian court-level metadata. Those provider metadata were not used to classify court levels. The preserved source text and official law link remain the basis of the stated voting rule.

Published judgments are a selected dataset

Lovdata describes its coverage as most appeal-court judgments and a selection of rulings and decisions, with free appeal-court decisions back to 2008. This is useful for close reading and developing coding rules. It does not establish complete or representative coverage of all defendants, attempted appeals or guilt determinations.

A study of published decisions must be labelled that way and handle duplicate judgments, multiple defendants, mixed outcomes, quashed decisions and later proceedings. Refused appeals may be selected out before any hearing. Published-decision rates cannot be presented as population acquittal rates without a documented coverage comparison.

A reproducible design

Request anonymous aggregate counts for the complete years 2019–2025, separately for each appeal court and nationally. The primary unit should be one defendant in one completed consideration of the guilt question. A person may occur in more than one case. Appeals confined to sentence, procedure or points of law must be distinguished from reconsideration of guilt.

Keep full acquittal, mixed outcome, conviction on all tried counts, other disposition and unknown outcome separate. Refusal of permission to appeal belongs to a different procedural stage. The definitions must be agreed with the data owner; sanctions alone are not always an unambiguous guilt indicator.

Full acquittal rate = full acquittals / (full acquittals + mixed outcomes + conviction on all tried counts)

This rate covers only cases with a classifiable determination of guilt. Other and unknown outcomes must remain visible alongside it. Missing counts are not zero. A 95% Wilson interval describes uncertainty under a specified binomial model; it does not correct missing cases, coding error, selection or dependence between repeated people and panels. For a genuinely complete enumeration, the observed period rate itself is exact; modelling intervals concern generalisation.

Sources and coverage

• Statistics Norway: sanctions, definitions and population
• Table 13725: court instance, without sex
• Table 13724: sex, without court instance
• Courts Administration annual report 2025, page 18
• Criminal Procedure Act, section 35
• Lovdata's stated judgment coverage

The checked catalogue and annual reports do not establish that the required data cannot exist internally. The data requests were sent to the Norwegian Courts Administration and Statistics Norway on 2 November 2026, asking them to clarify definitions, coverage and possible delivery.